CNBC Television · Sep 29, 2026
With Sam Altman, CEO of OpenAI
OpenAI did not sign on to Nvidia's new AI-safety guardrails software, even though rivals including Anthropic did.
Nvidia CEO Jensen Huang has framed AI safety as an engineering problem his software can fix. Altman, whose company was the one Huang said the guardrails would have prevented from a recent incident involving Hugging Face, argued that framing misses something.
"I don't think it's a full solution and I worry that if we treat AI safety as only an engineering problem, we will miss the very important point that we have a science problem in front of us."
Altman spoke to CNBC's Kate Rooney from the floor of OpenAI's own DevDay, the conference where the company was making its pitch to the developers, companies and investors it needs to keep building on its platform.
The full segment is covered here so you can skip it.
Here are the 6 takeaways that matter.
Key Takeaways
OpenAI did not join Nvidia's new AI-guardrails project, though Altman called it "a good thing"
He said OpenAI is doing similar work, but that treating safety as only an engineering problem misses "a science problem"
He said AI liability will likely split across several parties, the way blame does after a car crash
Faulty parts, a manufacturer's assembly error and a drunk driver can each carry the fault in that analogy
He said no open-source model beats OpenAI's on a cost-per-price comparison at any tier
Asked if Zuckerberg's Muse worries him, Altman said no and pointed to OpenAI's own hardware plans
He teased new OpenAI hardware without giving a timeline: "something that's worth waiting for"
Being mocked in a movie trailer about him doesn't bother Altman, he said
1. Nvidia's New Guardrails
Kate Rooney asked Altman whether OpenAI draws outsized scrutiny compared with rivals, then turned to the news: Nvidia had released guardrails software that Huang said would have prevented the Hugging Face incident, an engineering fix that Anthropic and other companies signed on to. Altman did not say why OpenAI stayed out of the project, but he answered Rooney's follow-up question of whether Huang was right that the problem could have been engineered away.
He said the guardrails are a real contribution, but not a fix for AI safety's underlying science problem
"I think it's a good thing. I think we are doing similar things. I don't think it's a full solution and I worry that if we treat AI safety as only an engineering problem, we will miss the very important point that we have a science problem in front of us." — Sam Altman
